ePI is coming to EU medicines. But digitising product information won't fix bad regulatory data.

ePI is not a PDF leaflet on a website

Electronic product information is often described as the digital future of the SmPC and package leaflet. That is technically correct, and it undersells the change.

EMA defines ePI as authorised, statutory product information for medicines — the summary of product characteristics, package leaflet and labelling — adapted for handling in electronic format. The format is the EU ePI Common Standard, based on HL7 FHIR, the interoperability standard used to exchange health data. EMA's stated aims are better accessibility, searchability, multilingual use and integration with electronic healthcare systems.

The difference from today's workflow is structural, not cosmetic. Most RA organisations still produce product information as a document and hand it on as one:

ePI is not simply a better PDF. It turns product information from a document into structured regulatory data — and that is an organisational change, not a formatting one.

Why this is a 2026 topic

ePI has been discussed for years, and a pilot ran from July 2023 to August 2024 with Denmark, the Netherlands, Spain and Sweden. In 2026 it moved from concept towards operations.

The submission guide covers when to create, update and submit ePI in the PLM Portal, how submission timelines align with procedure types, and how ePI is handled on suspension, withdrawal or refusal. That is procedural detail for a live process, not a vision document.

EMA's work is aligned with the reform of the EU pharmaceutical legislation, which addresses electronic product information. What becomes mandatory, for which products and from when, depends on the final texts and their implementation — see our article on the reform.

The testing call is itself a signal. EMA asked for people experienced in medicines information, product lifecycle management or labelling operations, ideally with structured-data and FHIR skills. That list of profiles is the list of functions ePI will touch inside a company.

Structured does not mean correct

A wrong sentence in a PDF has a limited reach. Someone reads it, and either notices or does not. A wrong value in structured data travels: to a website, an app, a hospital system, a search index, and increasingly to AI systems that summarise it for someone else.

1. An error enters the structured product information.
2. It is published through the data interface.
3. Websites and apps render it for patients.
4. Healthcare systems ingest it for prescribers and pharmacists.
5. Downstream tools, including AI, repeat it in new contexts.

Digitalisation increases the value of good data. It can also increase the reach of bad data.

This is not a hypothetical concern. Product data already held in structured form has known quality issues. In July 2026 EMA ran a webinar on managing product data quality in PMS, its Product Management Service, covering processes, known data-quality issues and best practice in areas such as MAH details, composition and indications. ePI adds the authorised text itself to that landscape.

FHIR solves structure. It does not solve truth.

FHIR can say that an element represents the therapeutic indication. It can make that element findable, translatable and exchangeable. It cannot know whether the content is right.

• Is this the currently authorised indication?
• Does this wording match the latest approved SmPC?
• Has a variation changed this section since it was structured?
• Is this translation aligned with the approved master text?
• Is this data consistent with what another regulatory system holds for the same product?

Those are regulatory questions, and they are answered by the regulatory record — approvals, procedure outcomes, versions — not by the data model. A well-formed resource that validates perfectly against the standard can still carry superseded text. Governance is what connects the structure to the truth.

This is not just a job for the labelling team

The natural reflex is to hand ePI to labelling. That is too narrow. ePI sits across product information and labelling, variations and lifecycle management, translation, regulatory operations, master data and the IT that holds structured content.

Which leads to the question most companies have not answered: who owns product information as data? Regulatory affairs owns the approved content. Labelling owns the artwork and the text process. Regulatory operations owns submissions. Master data teams own product records. IT owns the platforms. When two of those disagree about a field, someone has to have the authority to decide — and that decision needs to be traceable to the regulatory record.

The transition may be harder than the destination

A mature ePI ecosystem could simplify regulatory operations: one approved structured source, many representations, fewer manual re-keying steps. Getting there is the hard part.

• Existing product information, authored as documents, has to be converted into structure.
• Every language version has to be structured and kept aligned.
• Current regulatory systems have to coexist with new PLM Portal workflows.
• Data quality has to be good enough to publish as data.
• Internal ownership has to be settled before, not after, the first structured update.
• Validation of structured content needs defined checks.
• Procedures change, and teams will run old and new processes in parallel.

Regulators are working through the same transition. EMA's NCA Open Day in June 2026 was about national preparedness and a phased implementation approach. Readiness will not arrive everywhere at once, and planning should assume a period in which paper, PDF and ePI all exist for the same products.

Structured product information could be very valuable for AI

This is a technical consequence, not a marketing point. Today, an AI system reading product information usually starts from a PDF: it extracts text, infers headings from layout, reconstructs tables and guesses which paragraph belongs to which SmPC section. Every step can lose meaning.

Properly structured ePI removes most of that guessing. Section 4.1 is known to be section 4.1. The relationship between SmPC and package leaflet content can be explicit. That makes several useful things far more reliable: cross-product comparison, consistency checking between SmPC, leaflet and labelling, change detection between versions, label-impact analysis after a variation, and retrieval that cites the exact section it used.

AI becomes more useful when regulatory information becomes structured. It also becomes more dependent on the quality and governance of that structure.

Example: a variation changes SmPC section 4.2

Today, an approved posology change is carried through documents: update the SmPC, update the leaflet where it follows, manage translations, and replace the PDFs in the dossier lifecycle. With ePI, the same approval has a longer tail.

1. The variation is approved.
2. The authorised content changes.
3. The ePI is updated in the PLM Portal.
4. The affected structured sections are identified — 4.2, and any leaflet section that mirrors it.
5. Translations and related content are updated.
6. The new version is published.
7. Downstream systems consume the new information.

Now the regulatory question: how do you prove that every downstream representation reflects the approved change, and only the approved change? For a PDF, the answer is the document. For structured data, it is a chain — approval, version, section, translation, publication — and each link needs evidence. That is a regulatory-data problem, and it is new for most teams.

The patient side

The benefit for patients and healthcare professionals is real. Current information that is easier to search, easier to update, available in more languages and reachable from the pack is better than a folded leaflet that may predate the last safety update.

But digital access only improves patient information if the information behind it is current, comprehensible and trustworthy. A QR code is not a patient-safety strategy by itself.

What RA teams should start asking now

None of these questions needs the final implementation details to be answered. All of them take longer than expected once the first structured submission is due.

1. Who owns structured product information internally?
2. Is regulatory product data consistent across the systems that already hold it?
3. How are approved product-information changes traced to downstream content?
4. Is there one controlled source of truth for approved text?
5. How are translations governed against the approved master?
6. Does the organisation understand the EU ePI Common Standard and its FHIR model?
7. Who will operate the PLM Portal workflow?
8. How will ePI updates be integrated with variation procedures and timelines?
9. How will structured content be validated before submission?
10. What happens — and who decides — when two systems disagree?

Our view: ePI is a data-governance project disguised as a document project

Don't measure success by the number of converted documents. Measure it by whether the information is consistent, current and traceable to the approval behind it.

Don't treat FHIR as the solution. It is an excellent structure and interoperability layer. It does not replace regulatory ownership of what the structure contains.

Don't automate before fixing the underlying data. When websites, apps, healthcare systems and AI all consume the same structured information, correcting an error before it propagates matters more, not less.

Why this matters to us at RafiHive

RafiHive was built around a simple idea: regulatory information is only useful when its source, status and context can be verified. ePI moves the industry towards more structured, machine-readable regulatory information, and we think that is a major opportunity.

But structured regulatory information is not automatically factual regulatory intelligence. The harder problem remains knowing which information is current, authoritative, applicable and supported by the regulatory record. That is the problem RafiHive's packaging and labelling check works on today: a cited structural check of the SmPC, labelling and package leaflet against the required particulars, the current QRD template and each target market's rules, for qualified review.